An Efficient Long-Context Ranking Architecture With Calibrated LLM Distillation: Application to Person-Job Fit

TL;DR

This paper introduces a novel long-context ranking model for person-job fit that efficiently handles lengthy, structured resumes using a new cross-attention architecture and LLM-based distillation, improving relevance and interpretability.

Contribution

It presents a new long-context re-ranking architecture combined with LLM distillation to improve person-job matching accuracy and interpretability with minimal computational cost.

Findings

Outperforms state-of-the-art baselines in relevance and calibration

Efficient handling of long, structured, multilingual resumes

Produces interpretable skill-fit scores

Abstract

Finding the most relevant person for a job proposal in real time is challenging, especially when resumes are long, structured, and multilingual. In this paper, we propose a re-ranking model based on a new generation of late cross-attention architecture, that decomposes both resumes and project briefs to efficiently handle long-context inputs with minimal computational overhead. To mitigate historical data biases, we use a generative large language model (LLM) as a teacher, generating fine-grained, semantically grounded supervision. This signal is distilled into our student model via an enriched distillation loss function. The resulting model produces skill-fit scores that enable consistent and interpretable person-job matching. Experiments on relevance, ranking, and calibration metrics demonstrate that our approach outperforms state-of-the-art baselines.